Stop Duplicate Writes Under Load

The problem

Concurrent requests, retries, and background workers often modify the same records at once. This leads to oversold inventory, duplicate payments, and corrupted data through read-modify-write races.

What it does

• Identifies race conditions like Check-Then-Insert and Read-Then-Update in your application code.
• Applies database-level unique constraints and conditional atomic updates to guarantee data integrity.
• Implements idempotency keys for webhooks and payment processing to prevent duplicate side effects.
• Generates deterministic concurrency tests that simulate multiple actors hitting the same resource simultaneously.
• Adds optimistic concurrency control using version tokens to prevent stale writes.

Frameworks & tools

Works with any database-backed application including SQL (PostgreSQL, MySQL), ORMs, background workers, and message queues.

Why this beats prompting it yourself

Generic prompts often suggest thread-unsafe in-memory locks or global locks that kill performance. This skill enforces database-level correctness, ensures transactions don't span slow API calls, and provides a structured output that proves the fix via actual concurrency simulations.

Use cases

• Preventing multiple users from booking the same reservation slot.
• Ensuring a coupon is only redeemed once per account during high-traffic sales.
• Deduplicating incoming webhooks from providers like Stripe or Shopify.
• Fixing race conditions in inventory counters and account balance updates.

Known limitations

Cannot undo external side effects that occurred before a database rollback. Requires a test environment to verify concurrency fixes through simulation.